A wavelet packets and PCA based method for testing of analog circuits

Abstract

A method based on wavelet packets and PCA is developed for testing of analog circuits. It can detect both hard and soft faults in an analog electronic circuit by analyzing its output voltage response. The wavelet packets decomposition tree of the output voltage response is computed and the energy of every decomposed signal is used to form the feature vector. These features are combined by PCA to detect faults of the circuits-under-test (CUT). The principal component model of fault-free circuits is constructed before the actual testing process. Then we can use this principal component model to test the CUT. The features of the CUT are compared with the principal component model to calculate the statistic for fault detection. The proposed method can overcome the difficulty in selecting features and determining thresholds according to the expert's empirical knowledge. It was used to test the Sallen-Key filter and compared with existing methods. The results show the effectiveness of the proposed method.